Bike Sharing Systems (BSS) have become one of the main components of urban transportation networks, providing convenient and sustainable alternatives for commuters. Accurate prediction of ridership is crucial for efficiently managing these systems, ensuring optimal resource allocation, and enhancing user experience. This paper explores various modeling approaches to predict the number of users in BSS, focusing on the Toronto BSS as a proof of concept. These foundational steps are essential for extracting meaningful insights from the raw data generated by BSS. Understanding the intricate relationships between temporal, environmental, and spatial factors enables the development of robust predictive models. Our study applies a diverse set of machine learning (ML) techniques, including linear regression, linear Poisson regression, random forest, XGBoost regression, XGBoost Poisson regression, and XGBoost Poisson regression with exposure. Each model brings a unique perspective to the task, leveraging its strengths in capturing complex patterns within the data. The analysis encompasses historical usage patterns, weather conditions, and geographical features to enhance the predictive capabilities of our models. By comparing and contrasting different approaches, we provide valuable insights into the strengths and limitations of each method. Results indicate varying degrees of predictive performance among the models, with XGBoost Poisson regression with exposure emerging as the most effective method for anticipating the number of users in the BSS.